Classification of Variants across paired longitudinal samples in the PBTA Cohort

Set up

suppressPackageStartupMessages({
  library(tidyverse)
})

Directories and File Inputs/Outputs

# Detect the ".git" folder -- this will be in the project root directory
# Use this as the root directory to ensure proper sourcing of functions 
# no matter where this is called from
root_dir <- rprojroot::find_root(rprojroot::has_dir(".git"))
analysis_dir <- file.path(root_dir, "analyses", "tmb-vaf-longitudinal")
results_dir <- file.path(analysis_dir, "results")
input_dir <- file.path(analysis_dir, "input")
files_dir <- file.path(root_dir, "analyses", "sample-distribution-analysis", "results")

# Input files
pbta_file <- file.path(files_dir, "pbta.tsv") # file from add-sample-distribution module
genomic_paired_file <- file.path(files_dir, "genomic_assays_matched_time_points.tsv")
tmb_vaf_file <- file.path(results_dir, "tmb_vaf_genomic.tsv")
palette_file <- file.path(root_dir, "figures", "palettes", "oncoprint_color_palette.tsv")

# File path to plot directory
plots_dir <-
  file.path(analysis_dir, "plots")
if (!dir.exists(plots_dir)) {
  dir.create(plots_dir)
}

source(paste0(root_dir, "/figures/scripts/theme.R"))

Read in data and process

# Let's count #samples per cancer groups and timepoints.
# We will use the cg_id col that indicates cancer type as identified at the first diagnostic sample
timepoint_cg_n_df <- df %>% 
  count(cg_id, tumor_descriptor) %>% 
  dplyr::mutate(tumor_descriptor_cg_n = glue::glue("{cg_id}_{tumor_descriptor}  (N={n})")) %>% 
  dplyr::rename(timepoint_cg_n = n) 

# Let's count #samples per cancer groups and timepoints 
timepoint_cgGFAC_n_df <- df %>% 
  count(cgGFAC, td_cgGFAC) %>% 
  dplyr::mutate(tumor_descriptor_cgGFAC_n = glue::glue("{cgGFAC}_{td_cgGFAC}  (N={n})")) %>% 
  dplyr::rename(timepoint_cgGFAC_n = n) 

# Create df to use for plots
df_plot <- df %>% 
  left_join(timepoint_cg_n_df, by = c("tumor_descriptor", "cg_id")) %>%
  left_join(timepoint_cgGFAC_n_df, by = c("td_cgGFAC", "cgGFAC")) %>% 
  filter(!timepoint_cg_n <= 2,
         !timepoint_cgGFAC_n <= 2,
         !cg_id == "NA") %>% 
  mutate(tumor_descriptor = factor(tumor_descriptor),
         tumor_descriptor = fct_relevel(tumor_descriptor, timepoints))

Warning: There was 1 warning in `mutate()`.
ℹ In argument: `tumor_descriptor = fct_relevel(tumor_descriptor, timepoints)`.
Caused by warning:
! 1 unknown level in `f`: Unavailable

Define parameters for plots

# Read color palette
palette_df <- readr::read_tsv(palette_file, guess_max = 100000, show_col_types = FALSE) 

# Define and order palette
palette <- palette_df$hex_codes
names(palette) <- palette_df$Variant_Classification

Warning: Unknown or uninitialised column: `Variant_Classification`.

# Define label for plots
Alteration_type <- df_plot$Variant_Classification

# Define ylim
ylim <- max(df_plot$log10_tmb)

What type of alterations we observe per tumor descriptor?

# Create bxp
print(ggpubr::ggboxplot(df_plot, 
                        x = "tumor_descriptor", 
                        y = "log10_tmb", 
                        color = "Variant_Classification",
                        palette = palette) +
        theme_Publication() + 
        scale_y_continuous(limits = c(0, ylim)) +
        xlab("Timepoint") +
        theme(axis.text.x = element_text(angle = 90)))

# Save the plot
ggsave(filename = "Alteration_type_timepoints.pdf", 
       path = plots_dir, 
       width = 15, 
       height = 8, 
       device = "pdf", 
       useDingbats = FALSE)

What type of alterations we observe per tumor descriptor in each cancer group?

# Create bxp
print(ggpubr::ggboxplot(df_plot, 
                        x = "tumor_descriptor", 
                        y = "log10_tmb", 
                        color = "Variant_Classification",
                        palette = palette) +
        facet_wrap(~cg_id) +
        theme_Publication() + 
        xlab("Timepoint") +
        scale_y_continuous(limits = c(0, ylim)) +
        theme(axis.text.x = element_text(angle = 90)))

# Save the plot
ggsave(filename = "Alteration_type_cg_id.pdf", 
       path = plots_dir, 
       width = 25, 
       height = 18, 
       device = "pdf", 
       useDingbats = FALSE)

What type of alterations we observe per tumor descriptor in each cancer group defined by cgGFAC?

df_plot_cgGFAC <- df_plot %>% 
  arrange(tumor_descriptor_cgGFAC_n)

# Create bxp
print(ggpubr::ggboxplot(df_plot_cgGFAC, 
                        x = "tumor_descriptor_cgGFAC_n", 
                        y = "log10_tmb", 
                        color = "Variant_Classification",
                        palette = palette) +
        theme_Publication() + 
        xlab("Timepoint") +
        scale_y_continuous(limits = c(0, ylim)) +
        theme(axis.text.x = element_text(angle = 90)))

# Save the plot
ggsave(filename = "Alteration_type_cgGFAC.pdf", 
       path = plots_dir, 
       width = 14, 
       height = 8, 
       device = "pdf", 
       useDingbats = FALSE)

cgGFAC_id <- as.character(unique(df_plot_cgGFAC$cgGFAC))
cgGFAC_id

[1] "ATRT"  "DMG"   "HGG"   "LGG"   "Other"

# Loop through variable
for (i in seq_along(cgGFAC_id)){
  print(i)
  df_sub <- df_plot_cgGFAC %>%
      filter(cgGFAC == cgGFAC_id[i])

  
   # Create bxp
  print(ggpubr::ggboxplot(df_sub, 
                        x = "tumor_descriptor_cgGFAC_n", 
                        y = "log10_tmb", 
                        color = "Variant_Classification",
                        palette = palette) +
        theme_Publication() + 
        labs(title = paste(cgGFAC_id[i])) +
        scale_y_continuous(limits = c(0, ylim)) +
        theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1)))
}

[1] 1
[1] 2
[1] 3
[1] 4
[1] 5

What type of alterations we observe per tumor descriptor in each cancer group (add _n))?

cg <- as.character(unique(df_plot$cg_id))
cg

 [1] "Embryonal tumor with multilayer rosettes" "Low-grade glioma"                        
 [3] "Ependymoma"                               "Medulloblastoma"                         
 [5] "Atypical Teratoid Rhabdoid Tumor"         "Diffuse midline glioma"                  
 [7] "High-grade glioma"                        "Ganglioglioma"                           
 [9] "Meningioma"                               "Pilocytic astrocytoma"                   
[11] "CNS Embryonal tumor"                      "Neuroblastoma"                           
[13] "Schwannoma"                               "Chordoma"                                
[15] "Malignant peripheral nerve sheath tumor"  "Choroid plexus carcinoma"                
[17] "Adamantinomatous Craniopharyngioma"       "Dysembryoplastic neuroepithelial tumor"  
[19] "Ewing sarcoma"                            "Rosai-Dorfman disease"                   
[21] "Neurofibroma/Plexiform"                   "Glial-neuronal tumor"                    
[23] "Hemangioblastoma"                         "Craniopharyngioma"                       

# Loop through variable
for (i in seq_along(cg)){
  print(i)
  df_sub <- df_plot %>%
      filter(cg_id == cg[i])
  
  # Create bxp
  print(ggpubr::ggboxplot(df_sub, 
                        x = "tumor_descriptor_cg_n", 
                        y = "log10_tmb", 
                        color = "Variant_Classification",
                        palette = palette) +
        theme_Publication() + 
        xlab("Timepoint") +
        labs(title = paste(cg[i])) +
        scale_y_continuous(limits = c(0, ylim)) +
        theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1)))
}

[1] 1
[1] 2
[1] 3
[1] 4
[1] 5
[1] 6
[1] 7
[1] 8
[1] 9
[1] 10
[1] 11
[1] 12
[1] 13
[1] 14
[1] 15
[1] 16
[1] 17
[1] 18
[1] 19
[1] 20
[1] 21
[1] 22
[1] 23
[1] 24

What type of alterations we observe per tumor descriptor in each cancer group and timepoint model?

df_plot <- df_plot %>%
      filter(!is.na(timepoints_models))

tm <- as.character(unique(df_plot$timepoints_models))
tm

[1] "Dx-Rec"     "Dx-Pro"     "Rec-SM"     "Pro-Rec"    "Dx-SM"      "Dx-Pro-Rec"

# Loop through variable
for (i in seq_along(tm)){
  print(i)
  df_sub <- df_plot %>%
      filter(timepoints_models == tm[i],
             !is.na(timepoints_models))
  
  # Create bxp
  print(ggpubr::ggboxplot(df_sub, 
                        x = "tumor_descriptor", 
                        y = "tmb", 
                        color = "Variant_Classification",
                        palette = palette) +
        facet_wrap(~cg_id) +
        theme_Publication() +
        ylab("TMB") +
        xlab("Timepoint") +
        scale_y_continuous(limits = c(0, ylim)) +
        theme(axis.text.x = element_text(angle = 90)))
}

[1] 1
[1] 2
[1] 3
[1] 4
[1] 5
[1] 6